I am not open to your ahistorical take on Google Chat and XMPP.

Google didn't do anything wrong by using an open standard.

They didn't do anything wrong by building a good interface that people liked to use.

And they didn't do anything wrong by disconnecting from the network when the spam and harassment outweighed the benefit to their users.

We, the XMPP community, failed to capitalize on success by diversifying the network. It's our own fault not enough nodes were there.

If you'd like to draw some conclusions about ActivityPub from this, it should not be that a network should disallow supernodes, but that we have to counterbalance them with a wide diversity of other nodes of different sizes with different value propositions.
That huge audience of GChat users was an immense asset, and we fumbled it.

@evan the should have worked with Google to fix the issue.

I've already seen a lot of people quick to push Threads off the fedi rather than deal with the issues it might create

@anclement @evan "Supernodes" are okay if they are not predators with a track record like meta. We know how meta runs its social networks and we ran away from its playforms for something different.

@dilmandila @evan but Meta brings eyeballs and users to the the fediverse. It also brings brands to the fediverse. I know a lot of people may not like those things.

But rather than block and run away, shouldn't we try and address the issues in the standard?

@anclement @dilmandila human beings aren't eyeballs.

Participation in the social web is a human right.

We will not be changing the ActivityPub spec to keep people trapped in walled gardens and out of the social web.

@anclement @dilmandila nobody should ever have to interact with anyone they don't want to on the social web. EVER. That is fundamental, and I will fight tooth and nail against any structure that doesn't let people shape their own social space.

But that is part of the protocol already. Blocking accounts and defederation already work, as well as blocking words. We need better support for Bayesian filtering, but I think that will come.